This is the Pytorch implementation for SIGIR 2020 paper:
SIGIR 2020. Xiangnan He, Kuan Deng ,Xiang Wang, Yan Li, Yongdong Zhang, Meng Wang(2020). LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation, Paper in arXiv.
Author: Prof. Xiangnan He (staff.ustc.edu.cn/~hexn/)
I just made a little modification for myself on the basis of https://github.com/gusye1234/LightGCN-PyTorch, for my own reading and reuse.
(Also see Tensorflow implementation)
(Also see PyTorch implementation)
The output of my model, I checked, is exactly the same as them.
run LightGCN on Gowalla dataset:
python main.py --decay=1e-4 --lr=0.001 --layers=3 --seed=2020 --dataset="gowalla" --topks="[20]" --embedding_size=64